August 2, 2017
Cloud Migration Realities
Almost half of all organizations in Nemertes' 2017/18 Cloud and Network Benchmark are using IaaS in production. However, on average they have only 12% to 17% of their workload in IaaS environments. Similar numbers use PaaS and at a similar rate. Moreover, based on our benchmark interviews with organizations large and not so large, most organizations large enough to have a data center do not intend to have all their work in the cloud anytime soon, if ever. Data centers will, it appears, continue to host around a third of the workload when things steady down in about 4 years.
We applaud this measured approach as realistic and practical. We see that on average, organizations save no money by shifting workloads to cloud. Why? They didn't take into account the architecture of the applications they moved. Taking a system that was not built to operate in a cloud-like way (dynamically scaling to minimize the amount of stuff in "always on" mode, using distributed databases, etc) and moving it naively into a cloud environment -- the old "lift and shift" --raises costs. It also can increase the complexity and difficulty (and therefore cost) of integrating, managing, and monitoring the services involved. Architecture drives storage requirements, network dependencies, and can affect system security and enterprise risk.
Workload Placement Process
So, to make the migration responsibly, organizations need to have a process in place to evaluate and prioritize workloads for movement to the cloud. Basically, they need a process to help identify, for each workload, the best place to host that workload. For us at Nemertes, this is a song we have been singing since about 2010! We might aptly title it "Born to Run" (with all apologies to The Boss) because proper workload placement is foundational to delivering continuity of services.
A process can be simple or complex, depending on the technology environment, IT staff, and overall company approach to IT. Whether simple or complex, it should take into account a broad set of considerations, including:
- Risk assessments. What risk is the organization exposed to if this application fails or is compromised? What is its role, operationally, and what data does it collect, hold, manage, or use?
- Storage cost. How much data will it generate, how long will it have to be retained, and how does it need to be backed up or replicated?
- Application performance. What factors control performance and how will different hosting options hinder or help optimizing on them?
- Application architecture. How well suited is the application to running cloud-style, with horizontal dynamic scaling, and a cloud database behind it instead of a DIY database?
- And...network performance, cost of data movement, cost of layered security services, and potentially many more.
Shockingly, only 53.8% of organizations with a workload placement process formally consider risk. Only 51% consider storage costs. As for the rest of these points, fewer than half of participants consider each.
So, the enterprise still has a long way to go in terms of achieving real maturity in cloud migration.
Make sure you have a cloud strategy guiding how you plan to integrate cloud into your service delivery portfolio.
Begin to evaluate the workloads in your data centers and colo facilities, to rank them for suitability and priority for migration to cloud.
Create a workload placement process so you evaluate based on a consistent set of criteria.
Include application architecture, risk, and cost as three key facets considered in the workload placement process, but don't stop there; add other criteria, informed by your strategy.
Use the resulting list to help drive your cloud migration and overall cloud roadmap.